Evaluation of interfractional shift corrections in Gamma Knife radiosurgery

The purpose of this study was to evaluate the ability of the GammaPlan treatment planning system used in the Gamma Knife Icon and Esprit machines to automatically correct large interfractional shifts in hypofractionated frameless treatments using a custom 3D printed phantom.

A custom 3D printed insert was created to modify a commercial head phantom. The water equivalence of polylactic acid (PLA) was determined by comparing optical densities of film using solid water and PLA as buildup. Lesions of varying size, shape and location were created on images of the phantom to replicate several clinical scenarios. Seven 5-fraction plans were generated using typical dose objectives. Six combinations of headrests/masks were created to simulate different setup positions. Masks were changed between fraction measurements to simulate interfractional shifts. The composite dose distributions of the corrected plans were measured using Gafchromic film and compared to the original dose distribution using Gamma Analysis.

The average deviation for Coverage, Paddick Conformity Index and Gradient Index across all plans remained unchanged between the corrected plan and the original plan. No deviation was greater than 0.01 between the three metrics. Compared to the original plans, the measured dose distributions produced an average Gamma Passing Rate of 98.3% for 3%/1mm, 99.6% for 2%2mm, 96.3% for 2%1mm and 92.7% for 1%/1mm. All but two dose distributions produced passing rates above 90%. For the dosimetric analysis of PLA, the optical densities at various MU values were found to be equivalent for solid water and PLA.

The correction algorithm was able to produce a corrected plan almost identical to the original plan in terms of plan statistics and dose distributions. The algorithm could accurately correct for extreme interfractional shifts beyond what is seen clinically. The results of this study also show the potential for PLA phantoms to be used as dosimetry tools.